基于CatBoost算法的P2P违约预测模型应用研究

在互联网金融背景下,智能算法的应用能够为经济领域提供新的思路和方向,有效推动中国互联网金融向“互联网+金融+智能”模式的转变。P2P网络借贷是通过P2P公司搭建的第三方互联网平台进行“个人对个人”的直接信贷。以人人贷平台为研究对象,运用特征工程技术,将CatBoost算法应用于构建P2P违约预测模型,并对违约影响因素进行综合分析。结果表明,CatBoost算法的预测准确率达96%,对实际结果的拟合效果较好,并能够对模型出错所导致的损失成本进行有效控制。此外,综合分析违约影响因素发现,借款人的信用情况对借款人违约行为影响较大,其中还清贷款次数、逾期次数与成功借款次数应作为借款人信用评估的重要参考指标。结合本文的研究成果与中国P2P行业发展状况,本文建议P2P平台积极促进数据分析与测算分析技术的革新与应用,政府及相关部门形成政策法规的同步发展,促成从平台内部到外部环境的合力发展态势。

Application Research of P2P Default Prediction Model Based on CatBoost Algorithm

In the context of Internet finance,the application of intelligent algorithms provide new ideas and directions for the economic field,effectively promoting the transformation of China's Internet finance to theInternet+Finance+Intelligencemodel.P2P online lending is aperson-to-persondirect credit through a third-party Internet platform built by P2P companies;Among them,Renrendai is a more representative and influential P2P online lending platform in China,which continues to operate steadily and has comprehensive user information collection.Take Renrendai platform as the research object,use feature engineering technology,innovatively applie CatBoost algorithm to construct P2P default prediction model,and comprehensively analyse the factors affecting the default.The results show that the prediction accuracy of CatBoost algorithm is up to 96%,which fits the actual results better,and can effectively control the cost of losses caused by model errors.According to the comprehensive analysis of the factors that affect default,the borrower's credit status has the strongest impact on the borrower's default behaviour,while the number of loan repayments,overdue and successful borrowing should be used as important reference indicators for the borrower's credit assessment.Combining the research results with the development of China's P2P industry,it suggests that the P2P platform should actively promotes the innovation and application of data analysis and measurement analysis technologies,the government and relevant departments form the synchronous development of policies and regulations,resulting in a joint development trend from the platform to the external environment.